Fiscal Policy and Labor Markets in Low Income Economies∗ Alan Finkelstein Shapiro† Tufts University
Nathalie Gonz´alez Prieto‡ International Monetary Fund
June 14, 2017
Abstract We study the implications of changes in distortionary fiscal policy on sectoral employment, unemployment, participation, and aggregate outcomes in low income economies. We propose a general equilibrium framework with search frictions that features: (1) endogenous labor force participation decisions; (2) a distinction between agricultural and non-agricultural self-employment; and (3) a feedback effect from public capital to private-sector productivity. We find that increasing taxes on salaried-firm profits and capital gains raises salaried employment, output, consumption, and unemployment, and reduces self-employment in non-trivial ways. Conversely, increasing consumption taxes raises self-employment and reduces salaried employment, output, and consumption. Abstracting from the presence of non-agricultural selfemployment and endogenous participation decisions—two margins of central importance low-income economies—imply minuscule employment and aggregate responses to plausible changes in fiscal policy. The prevalence of self-employment—a reflection of the level of development—plays an important role for the quantitative consequences of policy. JEL Classifications: J40, O10, E6, H2 Keywords: Low income economies, fiscal policy, labor markets, informality, search frictions
∗
The views in this paper are solely the responsibility of the authors and should not be interpreted as representing the views of the International Monetary Fund or the countries it represents. Any errors are our own. † Corresponding author. E-mail: Alan.Finkelstein
[email protected]. Correspondence: Department of Economics, Tufts University, Braker Hall, 8 Upper Campus Road, Medford, MA 02155. ‡ E-mail:
[email protected]. Correspondence: Development Macroeconomics, Research Department, International Monetary Fund, 700 19th St NW, Washington, D.C. 20431.
1
Introduction
Low-income countries (LICs) are well known to have higher rates of labor force participation, agricultural and non-agricultural self-employment, and informal employment relative to other income groups (see Section 2; De Vreyer and Roubaud, 2013; Gutman, Sy, and Chattopadhyay, 2015; Golub and Hayat, 2015).1 These features are particularly prevalent in Sub-Saharan African (SSA) economies. Indeed, informal employment in this region represents, on average, more than 80 percent of total employment, with self-employment accounting for the bulk of both informality and total employment (OECD, 2009). At the same time, SSA economies face substantial public infrastructure needs and, tied to the high rates of informality and self-employment, severe challenges in raising domestic revenue to finance public investment without relying on natural-resource revenue and foreign grants (see Tables 1 and 2 in Section 2; IMF, 2014, 2015a, 2015b). Creating the necessary domestic fiscal space is all the more pressing since this much-needed public investment can have nonnegligible positive effects on private-sector productivity. However, little is known about the labor market and aggregate consequences of moving towards a more dependable domestic fiscal structure in economies where self-employment, informality, and agricultural employment is highly prevalent. More broadly, understanding the channels through which fiscal policy changes influence sectoral and aggregate employment is critical to enacting effective policies, especially in economies where labor is the main source of income for the majority of households. In this paper, we present a general equilibrium model with frictional unemployment and productive public investment that embodies key features of the employment and tax structure in LICs, with a focus on SSA economies. Calibrating the model to Tanzania, a SSA economy with detailed and quality labor market data, we explore the impact of changes in domestic distortionary taxation on: (1) the allocation of sectoral employment 1
In the rest of the paper, we use the terminology low-income economies (LICs) to refer to the group of countries comprised of low- and lower-middle-income economies. LICs, as defined by the World Bank, are countries with a per capita gross national income below USD$1,050 as of 2014. Lower middle-income economies are countries with a per capita gross national income between USD$1,050 and below USD$4,130 (data.worldbank.org/news/new-country-classifications-2015). Most low-income and lower-middle-income economies have the highest rates of informal employment out of all income groups, and this holds regardless of how informality is defined (lack of access to pension benefits, health benefits, labor protections, etc...).
1
across key categories in SSA; (2) unemployment and labor force participation; and (3) key macroeconomic aggregates. The model allows us to shed light on the mechanisms through which changes in exogenous fiscal policy affect labor market and aggregate outcomes in a LIC context.2 Our framework is comprised of two production sectors—agriculture and non-agriculture— and three employment categories—agricultural and non-agricultural self-employment, and non-agricultural salaried employment. The non-agricultural sector is comprised of salaried and self-employed (one-person) firms. The inclusion of agricultural employment allows us to: (1) characterize the effects of policy on total unemployment and labor force participation; and (2) determine whether fiscal policy also affects the reallocation of sectoral employment across firm categories. We assume that only salaried firms—firms that account for a very small share of employment but, in relative terms, contribute substantially more to total output—are subject to capital gains and profit taxation. In addition, non-agricultural consumption is subject to consumption taxes. These distortionary sources of revenue account for the bulk of government revenue in several low– and lower-middle-income economies in SSA (see Section 2). Following related literature and given the importance of public investment in SSA economies, we assume a positive feedback effect from tax collection to public investment and private-sector productivity.3 Finally, in contrast to most developing-country labor search models, which tend to focus on salaried employment, we introduce endogenous sectoral labor force participation. Allowing households to reallocate their members across the different employment categories that are prevalent in LICs provides a key margin of flexibility that generates non-trivial employment and aggregate effects in the presence of productivity-enhancing public investment. As such, we argue that this margin is critical for characterizing the impact of policy changes in SSA economies. Our policy experiments consist of an empirically plausible change in fiscal policy that increases taxes on salaried capital gains and profits or taxes on non-agricultural consumption (the increase in either tax is the same in percentage-point terms). Our main findings 2
A natural second step would be to consider optimal (Ramsey) fiscal policy in such a context. However, understanding the basic mechanisms through which policy operates—regardless of whether these policies are optimal or not—is a necessary first step before undertaking a formal analysis of optimal policy, which we leave for future work. 3 IMF (2015b) offers evidence on the link between public investment and productivity.
2
are as follows. In the presence of productivity-enhancing public investment and high nonagricultural self-employment shares—a well-known feature of LICs—an exogenous increase in the tax rate on salaried profits and capital gains increases salaried employment, total output and consumption; it reduces non-agricultural self-employment in a non-negligible while leaving agricultural self-employment virtually unchanged; and it generates a modest increase in unemployment. Intuitively, the increase in tax revenue is channeled to productive public investment, which boosts salaried (and total) labor productivity. The rise in the latter increases salaried labor demand, which pushes households to reallocate their search efforts away from non-agricultural self-employment and into (non-agricultural) salaried employment, even if productivity is higher across the board. Of note, the positive reallocation effects towards salaried employment take place if (1) households can explicitly reallocate individuals searching for non-agricultural self-employment towards searching for salaried employment (that is, if there is an endogenous participation margin for all employment categories), and importantly (2) there is initially a large share of individuals in non-agricultural self-employment, which represent a readily-available pool of potential labor that salaried firms can tap into. As such, differentiating between agricultural from non-agricultural self-employment is crucial as agricultural self-employment responds only marginally to policy (changes in relative output prices are quantitatively minimal given plausible fiscal policy changes). Conversely, a commensurate increase in the tax rate on non-agricultural consumption affects households’ decisions between agricultural and non-agricultural consumption, but also the costs of labor force participation across all employment categories. As a result, and consistent with standard Walrasian models, an increase in consumption taxes increases the marginal cost of labor force participation and directly affects labor supply (or participation) decisions. In particular, all else equal, the rise in the cost of participation exerts upward pressure on labor costs and reduces salaried labor demand, which pushes households to reallocate their members towards self-employment to counteract the fall in salaried employment demand. This occurs despite the rise in participation costs across the board (that is, households are more inclined to move their members to employment states that become more accessible and less costly in relative terms). In equilibrium, higher consumption taxes result in lower salaried employment and output, and therefore in lower tax revenue from salaried 3
firm profits and capital; lower unemployment, consumption, and output; and in higher selfemployment and aggregate participation. Changes in public (productivity-enhancing) capital are minimal despite the increase in consumption taxes since, quantitatively, the increase in consumption-tax revenue is fully offset by lower capital-tax revenue due to lower salaried capital gains and profits as a result of the reallocation of individuals into self-employment. Importantly, while the changes in aggregate labor force participation and unemployment in response to either policy are small, the sectoral compositional changes in participation and employment are non-negligible. Thus, our analysis suggests that focusing on aggregate labor market measures alone may yield an incomplete picture regarding the employment effects of fiscal reform. To further stress the relevance of our framework and results for LICs, we show that from a quantitative perspective, the positive effects from capital taxes are decreasing in the economy’s level of development as reflected in the prevalence of self-employment in the economy. Put differently, conditional on having tax-funded productive public investment, the gains from raising revenue via distortionary capital taxation are increasingly limited as LICs become more developed. Furthermore, a comparison of our model to simpler alternatives illustrates how (1) the distinction between agricultural and non-agricultural self-employment and (2) the inclusion of endogenous labor force participation for all employment categories— both of which are generally absent in existing models—are important for the impact of fiscal policy changes. Specifically, the response of sectoral employment, unemployment, consumption, and output to higher distortionary taxes are (1) either minuscule when we abstract from endogenous participation or non-agricultural self-employment (a key employment state in LICs) in the economy. Our findings point to two important implications for the analysis of fiscal policy in LICs. First, despite accounting for a large share of total employment, agricultural self-employment appears to play a limited role as a relevant adjustment margin to plausible changes in fiscal policy. This stands in contrast with non-agricultural self-employment, which plays a central role. Second, even though total unemployment and aggregate participation seem to be relatively unresponsive to policy, there are non-negligible sectoral employment reallocation effects. Of note, the fact that labor supply decisions appear to be relatively more influential 4
than labor demand in a LIC context is consistent with self-employment—which is driven by households’ labor supply decisions—being by far the most prevalent source of employment in these economies. Despite this fact, most of the theoretical literature on frictional labor markets in LICs has generally abstracted from endogenous labor force participation and an explicit distinction between agricultural and non-agricultural self-employment. As such, our results stress the relevance of explicitly taking into account households’ endogenous sectoral labor supply and participation decisions jointly with the full set of available employment options that households have access to in the quantitative analysis of fiscal policy in LICs. Relative to existing literature, which we discuss in more detail below, our work is the first to characterize the effects of changes in exogenous fiscal policy on sectoral employment, unemployment, participation, and aggregate outcomes in a LIC context with equilibrium unemployment and labor force participation. Our model introduces a distinction between informal employment (reflected in self-employment) and unemployment; it explicitly incorporates endogenous participation for all employment categories, which proves to be critical for generating non-negligible employment responses to plausible policy changes; and it introduces productivity-enhancing public investment to analyze the effects of fiscal policy on sectoral employment, unemployment, and macro outcomes in economies. More broadly, the distinction between self-employment and unemployment is important for economies where self-employment rates are on average more than 70 percent of total employment, as is the case in SSA, and also in determining the potential differential effects of fiscal policy on unemployment vis-`a-vis informal (self-)employment.4 The rest of the paper is structured as follows. Section 2 presents a series of stylized facts on the labor market and fiscal structure of LICs in SSA that motivate our theoretical framework. Section 3 presents the model. Section 4 outlines the baseline calibration of the model. Section 5 presents the main quantitative results. Section 6 concludes. 4
While self-employment rates tend to be high in middle-income developing countries as well, these rates are much lower relative to both SSA and other LICs (OECD, 2009).
5
2
Related Literature and Stylized Facts
2.1
Related Literature
Our work is related to the search and matching literature on informality in developing economies (Ulyssea, 2010; Kerr, 2012; Charlot, Malherbet, and Terra, 2015), the literature that incorporates self-employment in general equilibrium Walrasian and search models (Gollin, 2008; Finkelstein Shapiro, 2014; Epstein and Finkelstein Shapiro, 2017), studies that consider self-employment and labor force participation in a context with equilibrium unemployment (Finkelstein Shapiro and Mandelman, 2016), and to the strand of work that incorporates rural or agricultural production and rural-urban flows (Harris and Todaro, 1970; Fields, 1975; Cook and Nosaka, 2005; Satchi and Temple, 2008, Zenou, 2011; Adam, Bevan, and Gollin, 2013; Gollin, Parente, and Rogerson, 2004; Restuccia, Tao Yang, and Zhu, 2008; Gollin and Rogerson, 2013).5 Using a search model with formal and informal salaried employment, Charlot, Malherbet, and Terra (2015) study deregulation in the goods market and changes in labor taxes in the context of Brazil. While they abstract from self-employment, they extend their baseline environment to allow for directed search. The latter adds a margin of flexibility that is related to the inclusion of endogenous participation in our framework. Our results suggest that the combination of endogenous participation and self-employment—and not endogenous participation alone—is important to characterize the quantitative effects of changes in fiscal policy in economies where self-employment is the prevalent source of employment. Closest to our economic environment are Cook and Nosaka (2005), Satchi and Tem5
Finkelstein Shapiro and Mandelman (2016) study the interaction of frictional self-employment and labor force participation and its impact on cyclical unemployment dynamics in an emerging-economy context with remittance fluctuations. Rud and Trapeznikova (2016) focus on wage dispersion and salaried employment in Sub-Saharan Africa in partial equilibrium search environment and argue that movements between salaried work and self-employment—modeled as home production—play an important role. Given the magnitude of self-employment and the high rates of participation in SSA, our preferred approach is to separate home production (effectively, captured by the population outside the labor force in our model) from self-employment. Gollin (2008) does not incorporate search and matching frictions, but his work highlights the importance and relevance of self-employment in low- and middle-income economies. Also, Kumar and Schuetze (2007) incorporate self-employment and employers in a partial equilibrium search model, but their application is not to developing countries. Finally, see Ag´enor and Aizenman (1999) for earlier work on informality and fiscal policy in a Walrasian developing country context, and Melina, Yang, and Zanna (2015) for a policy application of general equilibrium models in a LIC context.
6
ple (2008), Kerr (2012), and Adam, Bevan, and Gollin (2013), where the first three use search-based frameworks. Cook and Nosaka (2005) present a three-sector model with search frictions in the (formal) tradable sector, and frictionless agricultural and nontradable (informal) sectors.There is no distinction between unemployment and informal employment and labor force participation is exogenous. Satchi and Temple (2005) focus on a middle-income developing economy and allow for rural-urban migration, where the rural sector is Walrasian and informality is reflected in self-employment. They allow self-employed individuals to actively search for formal employment; their framework also implies no distinction between unemployment and self-employment. While the presence of endogenous search by the selfemployed introduces a margin of adjustment similar to endogenous labor force participation in our environment, their model differs fundamentally from ours since endogenous sectoral participation in our framework establishes an important linkage across all employment categories (and therefore sectors); that is, allowing for endogenous search by the self-employed without also allowing for endogenous salaried search can yield dramatically different results, as we show in our work. Moreover, their analysis and implications from corporate tax changes is different from ours since we focus on changes in tax revenue that are specifically reflected in (productive) public investment. This component is absent in both Cook and Nosaka (2005) and in Stachi and Temple (2008), and is relevant to include in economies with large public infrastructure needs that aim to implement fiscal restructuring in order to bolster public investment. Kerr (2012) proposes a model of Tanzanian urban labor markets to explore the link between wage work and self-employment and earnings variations across the two employment categories, but abstracts from considering non-urban employment, public investment, or fiscal policy. Finally, Adam, Bevan, and Gollin (2013) use a Walrasian general equilibrium environment with public investment and rural-urban transitions to explore the role of different policies, including taxes, on poverty in the context of Tanzania. Our work complements theirs by, among other things, incorporating endogenous participation decisions and focusing on fiscal policy changes, public investment, sectoral employment reallocation, and more importantly unemployment and labor force participation in a LIC environment with
7
equilibrium unemployment.6
2.2
Stylized Facts
This subsection presents evidence on the structure of labor markets, the state of public infrastructure, and the structure of government revenue in low income economies in SSA. Figure 1: Labor Market Characteristics in SSA LICs vs. Country Groups Value Added Agric. High Income Upper Middle Income Zambia Senegal Guinea Uganda Tanzania Rwanda Ethiopia
Participation Rate Upper Middle Income High Income Guinea Uganda Senegal Zambia Ethiopia Rwanda Tanzania
0
10
20 30 40 % of GDP
0 20 40 60 80 100 % Total Pop. Ages 15−64
Self Employment
Unemployment
High Income Upper Middle Income Senegal Tanzania Rwanda Zambia Uganda Ethiopia Guinea
Rwanda Guinea Tanzania Uganda Ethiopia High Income Upper Middle Income Senegal Zambia 0 20 40 60 80 100 % of Total Employment
0
5 10 15 % of Labor Force
Source: World Bank World Development Indicators. Note: Latest observation for each country. We included the countries that have information for all the indicators.
Labor Markets and Competition from Informal Sector Figure 1 shows that, relative to middle- and high-income economies, the contribution of agriculture to GDP, the contribution of self-employment to total employment, and the labor force participation rate are all considerably higher in low- and lower-middle-income economies in SSA. Also, despite 6
A related macro literature studies monetary policy in SSA and other developing economies, but abstracts from incorporating a labor market structure consistent with these economies (see, for example, Buffie et al., 2008, or Anand, Prasad, and Zhang, 2015).
8
the absence of safety nets, unemployment rates are non-negligible for several countries in the region.7 Importantly, while self-employment accounts for a very large share of employment in these economies, self-employment outside of agriculture accounts for close to 50 percent of non-agricultural employment. In contrast, non-agricultural self-employment represents around 12 and 30 percent, respectively, in developed and other developing countries. At the same time, data from the World Bank Enterprise Survey suggests that more than 70 percent of firms in the formal sector cite competition from unregistered firms as a challenge. This implies that micro and small firms, most of which are informal, operate in very similar markets and compete by providing highly substitutable goods and services relative to larger (formal) firms.8 Infrastructure Deficits and Revenue Shortfalls The contribution of public infrastructure to growth and productivity has been recently highlighted by Calder´on and Serv´en (2014), among others. As illustrated in Table 1, compared to other LICs outside of Africa, LICs in SSA face severe public infrastructure challenges. In turn, Figure 2 shows that, despite having relatively high public investment-GDP ratios, SSA LICs rely more heavily on grants, have lower tax revenue from corporations, and have lower revenue and expenditures shares relative to both emerging economies and advanced economies.9 Projections for 2021 suggests that the gap in tax revenue and expenditures between SSA and emerging and advanced economies will remain substantial despite efforts to boost revenue and efficiency, with a reduction in the share of revenue from grants.
7
For evidence on job creation and destruction in Sub-Saharan Africa, see Shiferaw and Bedi (2009). Specifically, the average share of formal firms (using data from 2006 to 2015, depending on the country) that cite competition from unregistered firms as a challenge in Burkina Faso, Burundi, Cˆote D’Ivoire, Kenya, Liberia, Malawi, Mali, Mozambique, Sierra Leone, Tanzania, and Uganda is 71.7 percent (data availability by country varies by year). Including more LICs strengthens this fact. 9 Note that a relatively high public investment share may simply reflect the relatively low levels of public infrastructure in these economies relative to other income groups. 8
9
Table 1—Infrastructure Challenges in Sub-Saharan Africa
Infrastructure Category Paved-Road Density (kms/100 squared kms of arable land) Total Road Density (kms/100 squared kms of arable land) Main-Line Telephone Density (per thousand pop.) Mobile Phone Density (per thousand pop.) Electricity Generation Capacity (MW per million pop.) Electricity Coverage (% of pop. with access)
SSA LICs 31
Other LICs 134
137
211
10 55 37 16
78 76 326 41
Source: Reproduced from Gutman, Sy, and Chattopadhyay (2015) (original source: Table 2.2 in Yepes, Pierce, and Foster, 2009). Notes: LIC refers to low-income countries. The category ”Other LICs” includes other low-income countries outside of Sub-Saharan Africa as categorized by the World Bank. The countries included in the Sub-Saharan Africa LIC category are: Burkina Faso, Burundi, Cˆote D’Ivoire, Liberia, Mali, Mozambique, Sierra Leone, Tanzania, Kenya, Malawi, and Uganda.
Figure 2: Revenue, Expenditure, and Taxes Across Income Groups: 2015 and 2021 Projections 2015
Percent of GDP
40 Advanced Economies Emerging Economies SSA LICs
30
20
10
0
Revenue
Expenditure
Grants
Taxes
Corp. Taxes Public Invest.
2021 Projections
Percent of GDP
40
30
20
10
0
Revenue
Expenditure
Grants
Taxes
Corp. Taxes Public Invest.
Source: IMF World Economic Outlook (WEO) April 2016. Notes: LICs refers to low-income countries. The countries included in the Sub-Saharan Africa LIC category are: Burkina Faso, Burundi, Cˆote D’Ivoire, Liberia, Mali, Mozambique, Sierra Leone, Tanzania, Kenya, Malawi, and Uganda.
10
Fiscal Structure and Tax Potential in Low-Income Sub-Saharan African Economies Taxes on goods and services, taxes on income, profits, and capital gains, and grants account for a substantial share of total revenue (see Table 2). Taxes on international trade and transactions are also relevant, but their importance varies considerably across countries. Importantly, several low-income and lower-middle-income economies in SSA face non-negligible gaps between their potential tax-to-GDP ratio and their current levels (Figure 3). This suggests that (1) there is further need to improve countries’ domestic fiscal structure, and (2) there is room to strengthen the domestic fiscal structure via fiscal reform in these economies. Table 2—Fiscal Structure in Low-Income SSA Economies
Source of Revenue
Percent of GDP
Taxes on Goods, Services Taxes on Income, Profits, and Capital Gains Grants
6.252 5.974
Percent of Total Revenue 28.81 27.44
4.280
18.47
Taxes on International Trade and Transactions Other Revenue Other Taxes Payroll Taxes Property Taxes Social Contributions
3.325
15.05
1.639 0.481 0.111 0.049 0.004
7.401 2.056 0.442 0.205 0.126
Source: IMF Government Finance Statistics (GFS) 2013. Notes: Other revenue includes property income, sales on goods and services, fines, penalties, forfeits, voluntary transfers other than grants, and unidentified revenue. The countries included are: Burkina Faso, Burundi, Cˆote D’Ivoire, Liberia, Mali, Mozambique, Sierra Leone, Tanzania, Kenya, Malawi, and Uganda.
11
Figure 3: Current Tax-to-GDP Ratios and Tax Potential in Low and Lower-Middle-Income Sub-Saharan African Economies (2014) 25
Current Tax−GDP Ratio Tax Potential
Percent of GDP
20
15
10
5
0
Camer. Ethiopia Gambia Kenya Madag.
Mali
Tanz. Uganda Zambia Camer.
Source: Reproduced from Figure 1.20, IMF (2015a) (IMF Regional Economic Outlook, Sub-Saharan Africa, October 2015). Notes: Kenya, Ghana, Cameroon, and Zambia are categorized as lower-middle-income countries, with the remaining countries being low-income economies.
Modeling Implications All told, the above facts suggest that: (1) agricultural and non-agricultural economic activity are both important in low income economies; (2) selfemployment accounts for a substantial share of agricultural, non-agricultural, and total employment; and (3) taxes on goods and services, capital gains, and profits, and grants represent the bulk of government revenue from distortionary sources.10 At the same time, most selfemployment activities are effectively excluded from direct taxation given their small scale. Moreover, these firms tend to compete directly with salaried (formal) firms (implying that the former supply highly substitutable goods and services relative to those supplied by the average salaried firm). Our model incorporates these features in an environment where we can study the effects of fiscal policy on sectoral employment, unemployment, labor force 10
We abstract from taxes on international trade since this would imply having a richer model of a smallopen-economy that exports and imports goods. Furthermore, tackling revenue from international transactions in Sub-Saharan Africa in a relevant and consistent way would also require taking into account the prevalence of informal cross-border trade in the region (see, for example, Golub, 2015). Given our focus on employment and not international trade per se, we leave these relevant issues for future work.
12
participation, and aggregate outcomes.
3
The Model
The economy is comprised of three agents—households, owners of salaried firms, and a government—and two production sectors—agricultural and non-agricultural. In turn, there are two categories of firms within the non-agricultural production sector—salaried firms (ultimately owned by salaried-firm owners) and self-employed firms (owned by households). Total non-agricultural output is given by the sum of output from salaried firms and output from non-agricultural self-employed firms. The agricultural sector is owned by households and comprised solely of self-employed (owner-only) firms. The government collects non-agricultural consumption taxes and salaried-capital-gains and profit taxes to finance (productive) public capital accumulation and lump-sum transfers to households.11 Public capital, which is taken as given by all firms and households, improves firm productivity for all firms.12 The assumption of productive public capital that bolsters private-sector productivity is consistent with existing studies (see Baxter and King, 1993; Hulten, 1996; Pritchett, 2000; and Esfahani and Ramirez, 2003; and Adam, Bevan, and Gollin, 2013). Salaried-firm owners consume non-agricultural goods and accumulate capital for salaried firms. These firms are the only ones subject to profit and capital-gains taxes, and they use capital rented from salaried firm owners and salaried labor supplied by households to produce.13 Hiring salaried labor is subject to standard search and matching frictions between households and firms, which gives rise to equilibrium unemployment. The non-agricultural self-employment sector is comprised of one-person firms created and owned by households. The creation and operation of these firms is time-consuming and entails a utility cost for 11
In practice, a segment of self-employed firms may purchase goods that are often produced by salaried (formal) firms, to be used as inputs in their production process, implying that these goods are subject to consumption taxes. At the same time, self-employed firms may form part of a production chain that culminates in final output that is ultimately sold by salaried firms. Our assumption regarding non-agricultural production is broadly and tractably consistent with this notion by assuming that firms outside of agriculture produce highly substitutable output, which may ultimately be subject to taxation. 12 Assuming that only salaried firms benefit from public capital does not change our main conclusions. A similar claim holds if we assume that firms benefit differentially from public capital. 13 Of note, assuming skilled and unskilled salaried labor, as in Ag´enor and Aizenman (1999), which can broadly be interpreted as having formal and informal salaried labor among salaried firms, does not change our conclusions. Results available upon request.
13
households.14 Each firm uses a single unit of labor (that is, these are owner-only or one-person firms). Similar to salaried firms, self-employed firms also benefit from productivity-enhancing public capital. Self-employed firms in the agricultural sector are similarly created and use an identical production technology to their non-agricultural counterparts.15 Following the search literature, a representative household has a large number of members who enjoy perfect consumption insurance within the household. The household consumes both agricultural and non-agricultural consumption goods and makes explicit sectoral labor force participation (labor supply) decisions for their members. Employed household members can be in one of three employment categories—salaried employment, and agricultural or nonagricultural self-employment. If not employed, they can be searching in a given employment category. Searchers are categorized as unemployed. Members who are neither employed nor searching are categorized as outside of the labor force. The total household population is normalized to 1. We abstract from modeling an explicit rural-urban margin as this would introduce additional complexity without necessarily providing additional insights.16 While the majority of agricultural employment is often located in rural areas, not all rural employment is in agriculture (see, for example, Henderson, Roberts, and Storeygard, 2013). Since our focus is on sectoral employment at the economy-wide level, we consider the rural-urban distinction to be less relevant, even though there may be non-negligible internal migration costs. The inclusion of these costs is unlikely to affect our main conclusions since households’ labor force participation decisions in our environment already entail (utility) costs. The latter can partially embody the costs of migration to search for employment.17 Salaried Matching Process Let vs,t be the (non-agricultural) salaried-firm vacancy postings. Let ss,t be the measure of (non-agricultural) salaried searchers within the household. Sectoral labor market tightness for salaried employment is given by θs,t ≡ vs,t /ss,t . 14
Allowing for resource costs for the creation of self-employed firms does not change our main conclusions. Assuming that the self-employment sector (in either agriculture or outside of agriculture) uses internallyaccumulated capital to produce does not change our results. See Table A4 in the Appendix. 16 For seminal work on rural-urban migration, see Harris and Todaro (1970). For recent search models that explicitly consider rural-urban migration, see Satchi and Temple (2008). 17 We discuss the implications of abstracting from a rural-urban distinction as part of our caveats at the end of the paper. 15
14
New salaried employment is created via a matching process between vacancies and salaried searchers. Following the search literature, the matching functions ms,t = ms (vs, ,ss,t ) is constant-returns-to-scale, increasing and concave in each of its arguments. The job-finding probability for salaried employment is given by p(θs,t ) = ms,t /ss,t and is increasing in market tightness, while the job-filling probability is given by q(θs,t ) = ms,t /vs,t . Households, Agricultural Production, and Non-Agricultural Self-Employment There is a representative household comprised of a large number of members who enjoy perfect consumption insurance within the household. Household members can be in three different employment states: employed in (non-agricultural) salaried employment ns,t , selfemployed in the non-agricultural sector ne,t , and self-employed in the agricultural sector na,t . They can also be in three search (or unemployment) states: searching for salaried employment, ss,t , or searching for non-agricultural or agricultural self-employment, se,t and sa,t , respectively. The economy-wide unemployment level is then given by: st = ss,t +se,t +sa,t . In turn, total labor force participation is given by lf pt = ss,t + se,t + sa,t + ns,t + ne,t + na,t = P j (sj + nj ) where j ∈ {s, e, a} . As such, the total unemployment rate is ut = st /lf pt .
The household’s problem is to choose consumption of agricultural and non-agricultural
goods cha,t and chn,t , the desired measure of household members employed in the (self-employed) agricultural sector na,t+1 , the desired measures of individuals working in salaried employment, ns,t+1 , and the desired measure of non-agricultural self-employment, ne,t+1 . In addition, they P i P∞ t h h h choose the measures of sectoral searchers sa,t , ss,t , se,t to maximize t=0 β u(ca,t , cn,t) − h j (sj + nj ) for j ∈ {s, e, a}, subject to the budget constraint
(1 + τ c )chn,t + pa,t cha,t = ws,t ns,t + pa,t za,t na,t (kg,t )αg + ze,t ne,t (kg,t )αg + Tt , and the perceived evolution of, respectively, salaried employment, non-agricultural selfemployment, and agricultural self-employment ns,t+1 = (1 − ρs ) [ns,t + ss,t p(θs,t )] ,
(1)
ne,t+1 = (1 − ρe ) [ne,t + φe se,t ] ,
(2)
15
na,t+1 = (1 − ρe ) [na,t + φa sa,t ] ,
(3)
where the function u satisfies u′ > 0, u′′ < 0 for each of its arguments. The function h captures the disutility from total labor market participation and satisfies h′ > 0, h′′ > 0 (Arseneau and Chugh, 2012). The left-hand-side of the household’s budget constraint is comprised of non-agricultural and agricultural consumption, where τ c is the (time-invariant) proportional tax rate on non-agricultural consumption goods.18 The right-hand-side includes income from having household members in salaried employment, ws,t ns,t , where ws,t is the real salaried wage. Households also receive income pa,t za,t na,t (kg,t )αg from agricultural selfemployment production ya,t = za,t na,t (kg,t )αg , where pa,t is the price of agricultural output relative to the price of non-agricultural output (normalized to 1), za,t is exogenous agricultural productivity, and kg,t denotes public capital, which the household takes as given. Following related studies that introduce public capital, we assume that 0 ≤ αg < 1. In addition, households receive income pe,t ze,t ne,t (kg,t )αg from total non-agricultural self-employment production ye,t = ze,t ne,t (kg,t )αg , where ze,t is exogenous non-agricultural self-employment productivity. Thus, public capital affects private-sector production in similar ways.19 Finally, Tt denotes lump-sum transfers from the government. Turning to the evolution of salaried employment, ρs and p(θs,t ) denote the exogenous separation probability and the endogenous household-side matching probability associated with working in salaried employment. In turn, ρe is the exogenous separation probability for self-employed individuals. Transitions from unemployment to self-employment are affected by the search efficiency of agricultural and non-agricultural self-employment searchers, which we assume to be fixed and given by 0 < φa and 0 < φe , respectively. The household’s first-order conditions yield an optimal choice over agricultural and nonagricultural consumption goods: ucha (cha,t , chn,t ) pa,t = , (1 + τ c ) uchn (cha,t , chn,t ) 18
(4)
Given our focus on different steady-state equilibria, we abstract from introducing time-varying taxes for simplicity. 19 Our main conclusions remain unchanged if we assume that the self-employment sector also uses internally-accumulated capital to produce. See Table A4 in the Appendix.
16
and three conditions that characterize the sectoral participation decisions in the non-agricultural and agricultural sectors: h′t (1+τ c ) ucn ,t
p(θs,t ) h′t (1+τ c ) ucn ,t
φe h′t (1+τ c ) ucn ,t
φa
= (1 − ρs )Ξht+1|t
ws,t+1 −
= (1 − ρe )Ξht+1|t
= (1 − ρe )Ξht+1|t
h′t+1 (1
h′t+1 (1
pa,t+1 za,t+1 (kg,t+1 )αg −
+τ )
ucn ,t+1
ze,t+1 (kg,t+1 )αg −
h′t+1 (1+τ c ) ucn ,t+1
c
+
p(θs,t+1 )
c
+τ )
ucn ,t+1 h′t+1 (1
+
ucn ,t+1
(5)
h′t+1 (1+τ c ) ucn ,t+1
φe
c
+τ )
,
+
,
h′t+1 (1+τ c ) ucn ,t+1
φa
(6)
,
(7)
where Ξht+1|t ≡ β uchn,t+1 /uchn,t . Intuitively, households equate the expected marginal cost of searching for employment (given by the marginal disutility of participating in the labor mar-
ket, or the left-hand-side of each of the three conditions) to the expected marginal benefit of searching. In the case of salaried employment, the marginal benefit is given by the Nash wage net of the disutility from working, plus the continuation value of the salaried employment relationship. Similarly, the marginal benefit of entering (agricultural or non-agricultural) self-employment is given by the marginal product of having an additional household member in self-employment net of the disutility cost from working, plus the continuation value of keeping a household member in self-employment next period. Importantly, these conditions imply that households can reallocate searchers across different employment categories in response to changes in fiscal policy. In particular, all else equal, changes in consumption taxes have a direct impact on the incentive to search for employment in any given employment category. Also, note that in contrast to self-employment participation decisions and as is standard in search models, salaried labor force participation decisions are also affected by firms’ labor demand decisions via wages and market tightness. Salaried Firm Owners Salaried firm owners own and rent capital to salaried firms. They choose non-agricultural and agricultural consumption ckn,t and cka,t , and capital accumulation
17
ks,t+1 to maximize
P∞
t=0
β t u(ckn,t, cka,t ) subject to
(1 + τ c )ckn,t + pa,t cka,t + is,t = (1 − τ k ) [Πs,t + rs,t ks,t ] ,
(8)
and the evolution of salaried capital ks,t+1 = (1 − δ)ks,t + is,t ,
(9)
where τ k is the (time-invariant) tax rate on capital gains and profits—henceforth referred to simply as the capital tax rate—is,t is private investment, rs,t is the rental rate on capital, and δ is the exogenous capital depreciation rate.20 Πs,t are salaried-firm profits and are taken as given by salaried-firm owners. The salaried firm owners’ first-order conditions yield an optimal choice over agricultural and non-agricultural consumption goods: ucka (cka,t , ckn,t ) pa,t = , (1 + τ c ) uckn (cka,t , ckn,t )
(10)
and a standard capital Euler equation augmented to include capital income taxation: 1 = Ξkt+1|t (1 − τ k )rs,t+1 + (1 − δ) ,
(11)
where we define Ξkt+1|t = β uckn ,t+1 /uckn,t as the salaried firm owners’ stochastic discount factor.
Salaried Production and Wage Determination Salaried firms operate in the nonagricultural sector. They choose salaried vacancies vs,t , desired salaried employment ns,t+1 , P k k and private capital demand ks,t to maximize ∞ t=0 Ξt|0 (1 − τ )Πs,t subject to Πs,t = zs,t F (ns,t , k s,t , kg,t ) − ws,t ns,t − ψvs,t − rs,t ks,t, 20
We could assume a small open economy by allowing salaried firm owners to hold foreign debt. In the absence of public capital, an increase in capital or investment taxes would push salaried firm owners to reduce private capital accumulation (with adverse consequences for salaried labor demand and output) and to hold more foreign debt.
18
and the firm’s perceived evolution of salaried employment ns,t+1 = (1 − ρs ) [ns,t + vs,t q(θs,t )] , where zs,t is exogenous salaried productivity. The production function F (ns,t, k s,t , kg,t) is constant-returns-to-scale in salaried employment and private capital, and kg denotes public capital. The latter is taken as given by salaried firms. The inclusion of public capital in private sector production embodies the notion that public investment improves salaried-firm productivity (see, for example, Baxter and King, 1993, and Adam, Bevan, and Gollin, 2013, among others). The firm’s wage bill is ws,tns,t and ψ is the exogenous flow cost of posting a vacancy. The salaried firm’s first-order conditions deliver a standard job creation condition and a capital demand condition: ψ = (1 − ρs )Ξkt+1|t q(θs,t )
zs,t+1 Fns ,t+1 − ws,t+1 +
ψ q(θs,t+1 )
,
(12)
and rs,t = zs,t Fks ,t ,
(13)
where Fns and Fks denote the marginal product of salaried employment and the marginal product of private capital, respectively. The job creation condition equates the expected marginal cost of posting a vacancy to the expected marginal benefit of doing so, where the latter includes the continuation value of employment relationships. Note that given the structure of firm ownership, in steady state capital taxes will not directly affect firms’ decisions to hire workers, but these taxes will indirectly affect job creation through their impact on public capital (and therefore the marginal product of capital) and private capital accumulation (refer to the salaried-firm owners’ Euler equation). Following the search literature, the real wage is determined via bilateral Nash bargaining between workers and salaried firms. One can show that the Nash wage is given by (see the
19
Appendix for more details): ws,t
′ ht (1 + τ c ) p(θs,t ) − 1 ψ = χ zs,t Fns ,t + + (1 − χ) , q(θs,t ) uchn ,t p(θs,t )
(14)
where χ ∈ (0, 1) denotes the bargaining power of salaried workers. Non-Agricultural Total Production, Government, and Market Clearing Nonagricultural output is given by the sum of salaried and non-agricultural self-employment output, yn,t = ys,t + ye,t . The government uses revenue to finance public capital accumulation kg,t+1 and transfers to households Tt . The government budget constraint is given by21 Tt + kg,t+1 − (1 − δ)kg,t = τ c chn,t + ckn,t + τ k [Πs,t + rs,t ks,t] ,
(15)
In turn, the sectoral resource constraints are given by yn,t = chn,t + ckn,t + is,t + ig,t + ψvs,t ,
ya,t = cha,t + cka,t ,
(16)
where ig,t = kg,t+1 −(1−δ)kg,t is public investment. Total output is given by yt = yn,t +pa,t ya,t .
4
Functional Forms and Parameterization
The household and salaried-firm-owner utility functions over consumption are given by h i 1 1−σ j j j φc j φc φc 1 + (1 − γc ) ca,t with σ > 1, 0 < γc < 1, φc < 1 for u(ca,t , cn,t ) = 1−σ γc cn,t P hP i1+1/φ µ j = h, k. In turn, the disutility from participation is h (s + n ) = (s + n ) j j j j j j 1+1/φ for j ∈ {s, e, a} and µ, φ > 0 (see, for example, Arseneau and Chugh, 2012). This specifi-
cation for the disutility of participation delivers an empirically-factual negative relationship between steady-state output levels and self-employment shares. The salaried production function is F (ns,t , k s,t , kg,t) = (ns,t )α (ks,t)1−α (kg,t )αg , where 0 < α < 1, 0 ≤ αg < 1. The matching function for salaried firms is Cobb-Douglas: ms,t = Ms (ss,t )ξ (vs,t )1−ξ . 21
Allowing for exogenous government spending and exogenous sources of revenue (grants and revenue from additional sources outside of the scope of the model, such as royalties from natural resources) does not change our main conclusions.
20
We calibrate the model to Tanzania. A period is a year. Following the macro literature in low income economies, we set β = 0.95, the capital depreciation rate to δ = 0.08, the relative risk aversion parameter to σ = 2, and the private capital share to α = 0.32.22 As a starting point, we impose αg = 0.10 (see, for example, Berg et al., 2013), which is consistent with the macro literature that incorporates public capital in private-sector production, and investigate how our results change with alternative values. We normalize the productivity of salaried firms zs to 1. We initially set parameter φc in the CES aggregator for consumption to 0.5 implying imperfect substitutability between consumption categories. Our main conclusions are unaffected by other reasonable choices for these parameters.23 Given the unavailability of data on vacancies for low income economies, we cannot estimate the matching function for salaried employment. As such, we follow the search literature and set the matching elasticity and the bargaining power of salaried workers to 0.5.24 We initially set the parameter for the elasticity of participation φ = 1, implying quadratic utility costs from participation, and test the sensitivity of our results to this value. We set ρs and ρe to be consistent with the transition probabilities computed using Tanzania’s labor force survey for 2012-2013. This yields ρs = 0.025 and ρe = ρa = 0.043. Finally, we set the baseline tax rates using information from the Tanzania Revenue Authority so that τ c = 0.18 and τ k = 0.30. We normalize the search efficiency parameters φa and φe to 1 without loss of generality. Of note, we experiment with alternative values as part of our robustness checks and find that our main conclusions remain unchanged (see Table A3 and the accompanying discussion in the Appendix). Calibrated Parameters The remaining parameters M s , γc , ψ, µ, za , ze , and T are calibrated to match particular targets using Tanzanian data. The calibration targets are: the share of self-employment in non-agricultural employment (0.7365, Tanzanian Labor Force Survey); a probability of finding salaried employment of 0.45 (Tanzanian Labor Force Sur22
Our main conclusions remain unchanged under a lower β = 0.90 or a lower α = 0.22, which are plausible values in certain low income economies. 23 Differences in φc only affect the sensitivity of sectoral employment to changes in τ c , but not to changes k in τ . However, assuming a much smaller degree of substitution for the consumption aggregator does not change our main conclusions. Results available upon request. 24 A lower bargaining power for workers relative to the matching elasticity implies that the elasticity of salaried employment to tax rates is lower relative to our benchmark results, but our main qualitative conclusions remain unchanged.
21
vey, 2012-2013); a cost of posting vacancies that represents a small fraction of total output (1 percent of output); the share of food consumption in total consumption (0.48); a total labor force participation rate of 0.85 (World Bank World Development Indicators); the share of agriculture in total output (0.333, World Bank Development Indicators); and the share of public investment in GDP (0.07, World Development Indicators). The resulting parameter values are: M s = 0.1092, γc = 0.7414, ψ = 0.0343, µ = 3.0900, za = 2.4306, ze = 0.6853, and T = 0.0352. Our calibration therefore implies non-negligible productivity differentials between non-agricultural self-employment and salaried production.25
5
Numerical Experiments
To determine the effect of changes in fiscal policy on sectoral employment, unemployment, participation, and macro aggregates, we increase each tax rate (τ c or τ k ) by two percentage points by holding the tax rate that does not change at its baseline value. The magnitude of the change in fiscal policy we consider is meant to be illustrative but is also within plausible values in the context of fiscal reforms (see IMF, 2015b). In addition, we note that all distortionary tax changes take place while holding exogenous transfers T at their baseline value since we want to take into account the interaction between changes in fiscal policy and public capital and not between fiscal policy and transfers to households. Our main findings remain unchanged if we allow transfers to endogenously adjust to changes in revenue. Throughout our experiments, we assume that the official tax rates τ c and τ k are also the effective tax rates and that there are no losses in government absorptive capacity. Our main conclusions would not change if we assume a plausible degree of inefficiency in public capital (though the quantitative positive effect of an increase in capital taxes would be smaller; see the Appendix for more details). In what follows, households’ and capital h φc φc i φ1c for owners’ total consumption cht and ckt is given by cjt = γc cjn,t + (1 − γc ) cja,t
j = h, k. 25
Assuming that agricultural-sector productivity is smaller does not change our conclusions.
22
5.0.1
Main Results
Table 3 shows the elasticities of different labor market and aggregate variables with respect to individual changes in τ c or τ k of the same magnitude (the elasticity of variable x with respect to τ j is denoted by εx,τj for j = c, k)). Therefore, the numerical results we present in Table 3 as well as in subsequent tables should be interpreted as percent changes in the variable of interest in response to a 1% increase in each tax rate.26 Table 3—Benchmark Model Under Alternative Distortionary Tax Regimes: Elasticities
Original τ N ew τ εns ,τ εne ,τ εna ,τ εu,τ εlf p,τ εlf pna ,τ εlf pa ,τ εy,τ εch ,τ εck ,τ εkg ,τ
τc 0.180 0.200 −2.3801 0.9364 0.0412 −0.0975 0.0499 0.0551 0.0412 −0.2312 −0.0745 −0.0885 0
τk 0.300 0.320 5.7417 −2.4368 0.0778 0.1942 −0.1371 −0.2669 0.0778 0.8853 0.2411 0.1983 3.8239
Notes: εx,τ denotes the elasticity of the steady state variable x with respect to policy denotes labor force participation in the non-agricultural (agricultural) sector.
τ. lf pna (lf pa )
Consumption Taxes A 1% increase in consumption taxes (recall that the other tax rate is held at its baseline value) leads to a reduction in salaried employment of 2.4% and to an increase in non-agricultural (agricultural) self-employment of roughly 0.94% (0.04%). Output falls by 0.23%, while households’ and capital owners’ consumption decrease by roughly 0.08%. Given the magnitude of the change in taxes and the sectoral reallocation that takes place, the change in total unemployment is very small. Intuitively, an increase in consumption taxes, all else equal, increases household members’ marginal cost of participating in the labor market (refer to the optimal participation decisions of the household). This is most relevant for salaried employment. Indeed, the initial fall in salaried labor supply as a result of the increase in consumption taxes pushes salaried firms to reduce capital and labor 26
A similar comment applies to all results presented in the Appendix.
23
demand, resulting in an equilibrium reduction in salaried employment and output and, ultimately, in total output. As a result, households shift the allocation of household members towards agricultural self-employment, where consumption taxes are not collected, but more importantly towards non-agricultural self-employment to counteract the reduction in salaried employment opportunities. This occurs despite the fact that, as a result of the increase in τ c , the participation cost has increased across the board for all employment categories, but the incentive to search for salaried employment is more adversely affected in relative terms given the response of labor demand. As such, households ultimately redirect their members to self-employment. The marginal fall in unemployment is mainly explained by a rise in total participation, where the latter is explained by the rise in self-employment and not salaried employment (in fact, salaried participation drops drastically). Importantly, while revenue is directed to public investment, the adverse impact of the tax change on salaried-firm profits and capital gains is large enough to offset any initial revenue gains from higher capital taxes that would otherwise bolster public investment. As such, public capital remains virtually unchanged. All told, we conclude that reasonable changes in consumption taxes that aim to bolster public capital can have negative labor market and aggregate consequences by generating a reallocation of labor that offsets the gains from additional revenue. Capital Taxes In principle, an increase in taxes on profits and capital gains, all else equal, reduces salaried firm owners’ expected return to capital (refer to the salaried firm owners’ capital Euler equation). However, to the extent that the additional revenue is devoted to accumulating public capital, which in turn contributes to firm productivity, the increase in public capital bolsters salaried firms’ labor productivity and incentive to demand private capital and labor. As a result, a 1% increase in capital taxes generates a non-negligible increase in salaried employment of roughly 5.7%. Importantly, in response to the rise in salaried labor demand, households reallocate their members away from searching for nonagricultural self-employment and into searching for salaried employment, resulting in a sharp fall in non-agricultural self-employment of 2.43%. The policy also generates an increase in total output of 0.89%. However, unemployment increases by 0.19%, mainly because self-
24
employment plays such a prevalent role as a source of employment in the economy (thus, a fall in self-employment all else equal increases unemployment) and labor force participation falls slightly (in turn, this puts upward pressure on the unemployment rate).27 Agricultural self-employment responds marginally to the change in tax rates. Of note, this takes place despite the absence of explicit sectoral resource-based reallocation costs often associated to movements out of agriculture and into other sectors. The fact that agricultural employment is not responsive to changes in taxes is consistent with evidence suggesting that changes in productivity are one of the main drivers of transitions out of agricultural employment (Diao, Harttgen, and McMillan, 2016).28 As we show further below, the strong reallocation of sectoral searchers by households plays an important role in eventually boosting salaried employment and reducing non-agricultural self-employment, despite the fact that aggregate participation changes marginally. We show below that, despite the latter, allowing for endogenous participation and endogenous salaried labor supply is critical for quantitatively characterizing changes in sectoral employment. Finally, while higher capital taxes would clearly have a negative effect on capital accumulation (and therefore labor demand) absent any productive public investment, channeling the extra revenue into productivity-enhancing public investment means that public capital increases by 3.8%. Given the presence of public capital in (salaried-sector) production, this ultimately explains the positive and non-negligible effect of the policy on non-agricultural employment, salaried employment, and output relative to a similar consumption-tax-based policy. We therefore conclude that not all fiscal policy changes are created equal. In what follows, we show that accounting for self-employment and households’ ability to reallocate their members across employment categories plays a critical role for characterizing the quantitative effects of fiscal reforms. 27
As shown in Table A3 in the Appendix, conducting the same policy experiment when the efficiency of search for self-employment is lower leads to a reduction in unemployment in response to higher τ k since salaried employment responds more forcefully to the change in taxes. Intuitively, since it becomes more difficult to move to self-employment relative to the benchmark calibration, a change in taxes generates a sharper reallocation towards salaried employment. 28 Indeed, changing the sectoral productivity parameters in the model confirms that such changes generate a non-trivial reallocation of employment in the economy, which includes large changes in agricultural selfemployment.
25
5.0.2
The Importance of Endogenous Participation and Salaried Supply
Table 4 illustrates how our results change when consider a model where we shut down endogenous labor force participation and households’ decisions over the measure of salaried searchers. In this alternative model, salaried employment is primarily driven by labordemand considerations while self-employment continues to be optimally chosen by households (since, by definition, self-employment is solely a labor supply choice).29 Of note, recall that self-employment continues to account for the bulk of total employment (salaried employment represents only 16 percent of total employment under the baseline tax rates). Table 4—Benchmark Model vs. Economy Without Participation and Salaried Labor Supply Margin Under Alternative Distortionary Tax Regimes: Elasticities
Original τ N ew τ εns ,τ εne ,τ εna ,τ εu,τ εlf p,τ εlf pna ,τ εlf pa ,τ εy,τ εch ,τ εck ,τ εkg ,τ
Benchmark
Model
τc 0.180 0.200 −2.3801 0.9364 0.0412 −0.0975 0.0499 0.0551 0.0412 −0.2312 −0.0745 −0.0885 0
τk 0.300 0.320 5.7417 −2.4368 0.0778 0.1942 −0.1371 −0.2669 0.0778 0.8853 0.2411 0.1983 3.8239
No or Salaried τc 0.180 0.200 −0.1046 0 0.058 0.1165 −0.0823 −0.1601 0.0467 0.0608 0.1127 −0.0025 1.4408
Participation Labor Supply τk 0.300 0.320 0.0362 0 −0.0213 0.1942 −0.1371 −0.2669 0.0778 −0.0398 0.1525 −0.0232 0.4935
Note: εx,τ denotes the elasticity of the steady state variable x with respect to policy denotes labor force participation in the non-agricultural (agricultural) sector.
τ. lf pna (lf pa )
In the absence of endogenous salaried supply and participation, increases in τ c and τ k have negligible effects on salaried employment and non-agricultural self-employment. Despite households’ optimal choices over self-employment, the latter becomes virtually unresponsive to policy when household members cannot be redirected to search for salaried employment because salaried labor supply is now exogenous. This exercise confirms that 29
Briefly, when we remove endogenous participation, we assume that the labor force, which is now exogenous, is normalized to one and that households choose the amount of household expenditures needed to send individuals from unemployment into (agricultural and non-agricultural) self-employment. A choice over these expenditures is needed in order to allow households to continue to make an explicit choice over self-employment (see, for example, Finkelstein Shapiro, 2014, for a similar idea with financial frictions among the self-employed).
26
households’ option to reallocate their members across all employment categories—including salaried employment—via explicit sectoral participation decisions that include salaried employment is important for characterizing the effect of changes in policy, with labor demand playing a secondary role.30 Why are endogenous participation and salaried labor supply so critical for generating quantitatively sizeable responses to fiscal policy changes? Our calibration implies a low unemployment level, consistent with Tanzanian data. Also, in the absence of endogenous participation, unemployment is given by u = 1 − ns − ne − na . If ne and na are not very responsive to policy, and households cannot change the measure of salaried searchers across sectors—recall that the latter are part of the unemployed in the benchmark model—then all else equal u will be little changed. In addition, if salaried firms have a difficult time attracting searchers from other sectors because households cannot easily reallocate their workers in the absence of endogenous sectoral (salaried) participation, market tightness (and the job-filling probability) will not change much either. These elements combined ultimately lead to a quantitatively-small equilibrium response in unemployment, salaried employment, and more broadly other macro variables. Finally, there is one more contrasting result with respect to our benchmark findings: note that, absent endogenous participation, a change in consumption taxes will not be fully offset by a sharp (and endogenous) reduction in revenue from capital taxes. This is due to the fact that salaried firms are not adversely affected by changes in search behavior on the part of household members in response to changes in τ c . This explains why εkg ,τ is positive when τ c increases in the model without endogenous participation, whereas it does not in the benchmark model. 5.0.3
The Importance of Non-Agricultural Self-Employment
Table 5 shows the results for a version of the benchmark model where non-agricultural self-employment is absent. The response of labor market variables to changes in taxes is surprisingly different relative to the benchmark model. In fact, looking at a change in τ k , 30
A version of the benchmark model with endogenous salaried and self-employment search but fixed participation yields similar qualitative (though smaller quantitative) results to those in Table 3, which confirms the relevance of households’ choices over all margins of employment
27
the size of the response is considerably lower, and the sign is the opposite, relative to the benchmark model. To understand why εns ,τ , εy,τ , εch,τ , εck ,τ , and εlf p,τ change signs in response to a higher τ k relative to our benchmark findings, note that even if higher taxes all else equal bolster the stock of public capital, salaried firms are unable to attract potential workers who were previously searching for non-agricultural self-employment since the latter employment category no longer exists. Moreover, the tax has little impact on the relative price of agricultural output. As such, the rise in taxes induces a fall in salaried labor demand, capital, and output, which eventually leads to a reduction in capital gains, salaried profits, and total output. Without households’ ability to move workers from self-employment into salaried employment within the non-agricultural sector, the increase in taxes causes the equilibrium amount of consumption and total output to fall. Table 5—Benchmark Model vs. Economy Without Non-Agricultural Self-Employment Under Alternative Distortionary Tax Regimes: Elasticities
Original τ N ew τ εns ,τ εne ,τ εna ,τ εu,τ εlf p,τ εlf pna ,τ εlf pa ,τ εy,τ εch ,τ εck ,τ εkg ,τ
Benchmark τc 0.180 0.200 −2.3801 0.9364 0.0412 −0.0975 0.0499 0.0551 0.0412 −0.2312 −0.0745 −0.0885 0
Model τk 0.300 0.320 5.7417 −2.4368 0.0778 0.1942 −0.1371 −0.2669 0.0778 0.8853 0.2411 0.1983 3.8239
No τc 0.180 0.200 −0.0108 − −0.0088 −0.0633 −0.0131 −0.017 −0.0088 0.2012 0.0562 0.0033 1.4621
Non-Agric. SE τk 0.300 0.320 −0.0742 − 0.1038 0.0284 0.0117 −0.0705 0.1038 −0.1406 −0.039 −0.0566 0.5849
Note: εx,τ denotes the elasticity of the steady state variable x with respect to policy denotes labor force participation in the non-agricultural (agricultural) sector.
τ. lf pna (lf pa )
Importantly, these results suggest that, while the inclusion of public capital does matter for the quantitative effects of changes in fiscal policy on sectoral employment, the presence of public capital alongside non-agricultural self-employment is critical as well. Indeed, bolstering public capital, which in turn increases productivity, will not have the desired effects if salaried firms cannot easily respond to the rise in public capital by hiring more labor outside of agriculture. Since tax changes have negligible quantitative effects on the 28
price of agricultural output, households have little incentive to redirect their members away from agricultural self-employment and towards salaried employment, and this further reduces salaried firms’ possibility to hire workers. More broadly, our results show that when it comes to changes in fiscal policy, differentiating between the two types of self-employment—two types that are equally important in LIC labor markets—is relevant for assessing the possible impact of changes in fiscal policy on sectoral employment. 5.0.4
The Importance of Employment and Sectoral Output Allocation
Our baseline calibration reproduces a share of self-employment in non-agricultural employment of 0.74, in line with the data for Tanzania and other LICs. To illustrate how the impact of policy depends on the economy’s level of development as reflected in the allocation of employment across employment categories and the contribution of agriculture to output, we now assume an economy that resembles emerging and developed economies in terms of (1) having a smaller share of self-employment, and (2) a smaller contribution of agriculture to GDP relative to our baseline (low-income) economy. Note that a lower share of agriculture in total output also implies a smaller share of agricultural self-employment, which is consistent with the allocation of employment in more developed economies. For purely illustrative purposes, we halve the share of non-agricultural self-employment in total non-agricultural employment and the contribution of agriculture to output relative to our baseline calibration for Tanzania, and then conduct the same fiscal policy changes. Specifically, we recalibrate the economy using these two new targets. The new calibration implies that the share of non-agricultural self-employment in the total labor force falls from 0.438 to 0.227, and the share of agricultural self-employment in the total labor force falls from 0.361 to 0.226. Importantly, an economy with lower self-employment also has a higher level of output and consumption, which is consistent with the negative cross-country relationship between the share of self-employment and the level of development.
29
Table 6—Benchmark Model vs. Economy with Smaller Share of Non-Agricultural Self-Employment and Agricultural Production: Elasticities
Original τ N ew τ εns ,τ εne ,τ εna ,τ εu,τ εlf p,τ εlf pna ,τ εlf pa ,τ εy,τ εch ,τ εck ,τ εkg ,τ
Benchmark τc 0.180 0.200 −2.3801 0.9364 0.0412 −0.0975 0.0499 0.0551 0.0412 −0.2312 −0.0745 −0.0885 0
Model τk 0.300 0.320 5.7417 −2.4368 0.0778 0.1942 −0.1371 −0.2669 0.0778 0.8853 0.2411 0.1983 3.8239
Model with τc 0.180 0.200 −1.1143 2.7211 0.046 −0.1384 0.0681 0.0749 0.046 −0.2795 −0.0955 −0.1254 0
Smaller ne , pa ya /y τk 0.300 0.320 1.5686 −3.9415 −0.0087 0.0949 −0.1142 −0.1467 −0.0087 0.7042 0.2293 0.1407 3.8239
Notes: εx,τ denotes the elasticity of the steady state variable x with respect to policy denotes labor force participation in the non-agricultural (agricultural) sector.
τ. lf pna (lf pa )
As shown in Table 6, the smaller is the share of self-employment, the lower is the positive (negative) impact of changes in τ k (τ c ) on salaried employment, consumption, and output. This suggests that the quantitative conclusions above are heavily dependent on the economy’s level of development and its associated allocation of employment across sectors, with the benefits of fiscal policy changes being greater for economies at lower levels of development.31 Intuitively, the difference in policy impact across the two scenarios traces back to the higher marginal product of public capital in a LIC relative to more developed economies. All told, the results in Table 6 suggest that, despite bolstering firm productivity via higher public investment, increasing government revenue via a higher τ k is less effective in raising output, consumption, and salaried employment in economies where salaried work is a more prevalent source of employment (and the agricultural employment and output shares are smaller). 31 The reduction in self-employment we engineer is also accompanied by an endogenous increase in public capital and total output relative to the baseline calibration for Tanzania. This is consistent with the fact that more developed economies have higher public investment and output levels (even though public investmentoutput ratios may be lower). Moreover, if we were to include inefficient public capital in the model (which we discuss below), the results in Table 6 would suggest that increasing τ k would have even smaller positive effects on salaried employment, output, and consumption in economies where non-agricultural self-employment and agriculture account for a smaller share of the labor force in the economy, respectively. See Table A5 in the Appendix.
30
Efficiency of Public Investment Efficiency losses in public capital and investment are well-known to afflict LICs (IMF, 2015b). To explore the impact of such losses on our results, we assume that only a fraction of accumulated public capital becomes productive capital in the private sector. Specifically, the production functions for salaried firms, non-agricultural s self-employed firms, and agricultural self-employed firms become ys,t = zs,t F (ns,t , k s,t , kg,t )= s αg s αg s αg zs,t (ns,t )α (ks,t)1−α kg,t , ye,t = ze,t ne,t kg,t , and ya,t = za,t na,t kg,t , where productive s public capital is given by kg,t = kg,t − ϕk (kg,t )ηk , ϕk > 0 and ηk ≥ 1. Importantly, ϕk (kg,t)ηk
is a pure resource cost that reflects public investment inefficiencies and therefore appears in the economy’s resource constraint.32 We initially set ηk = 1 and calibrate ϕk such that this resource cost absorbs 40 percent of public investment, which is in line with evidence for LICs (IMF, 2015b). Table 7—Benchmark Model vs. Model with Inefficiency in Public Investment
Original τ N ew τ εnf ,τ εne,τ εna ,τ εu,τ εlf p,τ εlf pna ,τ εlf pa ,τ εy,τ εch ,τ εck ,τ εkg ,τ
Benchmark τc 0.180 0.200 −2.3801 0.9364 0.0412 −0.0975 0.0499 0.0551 0.0412 −0.2312 −0.0745 −0.0885 0
Model τk 0.300 0.320 5.7417 −2.4368 0.0778 0.1942 −0.1371 −0.2669 0.0778 0.8853 0.2411 0.1983 3.8239
Public Inv. τc 0.180 0.200 −2.2591 0.8884 0.041 −0.0926 0.0479 0.052 0.041 −0.219 −0.0727 −0.0872 0
Inefficiency τk 0.300 0.320 3.0155 −0.9469 −0.091 0.0815 0.03 0.1031 −0.091 0.7159 0.1561 0.1047 3.8239
Note: εx,τ denotes the elasticity of the steady state variable x with respect to policy τ. εkg ,τ and εkgs ,τ are quantitatively identical. lf pna (lf pa ) denotes labor force participation in the non-agricultural (agricultural) sector.
Table 7 shows that, as should be expected, allowing for inefficient public investment reduces the positive quantitative effect of changes in τ k on salaried employment, consumption, and output. Interestingly, despite having non-negligible efficiency losses in productive public capital, increasing τ k continues to have positive effects via the effect of higher productive public capital on salaried firms. As such, our qualitative conclusions remain unchanged 32
Alternative specifications of public capital and/or investment inefficiency yield similar conclusions.
31
under plausible parameterizations.33 The main takeaway from this experiment is that, despite having non-negligible losses in public investment efficiency, economies that have an allocation of employment and agricultural output consistent with LICs may still see positive effects on salaried employment, consumption, and output (with small quantitative effects on aggregate unemployment and participation) from increasing capital taxes in an environment with productivity-enhancing public investment. Finally, as shown in Table A5 in the Appendix, such benefits will be smaller once the economy is more developed (as reflected by smaller self-employment and agricultural output shares), unless the loss in public investment efficiency decreases as the economy becomes more developed. 5.0.5
Robustness Experiments and Some Caveats
We present and discuss several robustness checks in the Appendix. These include, among others, using alternative functional forms for the disutility of participation (Table A1 in the Appendix), and adopting alternative plausible calibrations (Tables A2 and A3 in the Appendix), allowing for capital in self-employment (Table A4 in the Appendix), and a comparison of our baseline economy and a more developed economy (in terms of having smaller self-employment and agricultural output shares) under public capital inefficiency (Table A5 in the Appendix). Some caveats are also in order. Evidence for LICs suggests that micro and small enterprises face significant capital and financial constraints relative to larger firms (Grimm et al., 2011). Our benchmark model abstracts from including such constraints in order to analyze the impact of fiscal policy changes on the labor markets in a transparent environment. Introducing capital constraints among self-employed firms, regardless of whether they operate in the agricultural or non-agricultural sectors, is unlikely to significantly affect our main results. Indeed, in the case of changes in τ k , where (non-agricultural) self-employment falls, the response of self-employment is likely to be larger since households would be reallocating searchers away from more constrained (smaller) firms into relatively less constrained (salaried or larger) firms.34 We note that the higher ηk is, the smaller are the employment and output benefits from increasing τ k , with values of ηk above 2 generating adverse effects from higher taxes. 34 This can be seen more clearly in Table A3, where a lower efficiency of search for the self-employed can 33
32
Turning to the reallocation process across firm categories and sectors, while our model establishes a distinction between agricultural and non-agricultural production, the cost of reallocation of searchers away from agricultural employment and into non-agricultural (selfemployed and salaried) employment is a utility cost purely associated to labor force participation, where the latter includes searchers and employed individuals for each employment category. The cost of search for salaried employment could be broadly interpreted as also embodying the cost of migration out of agriculture. Introducing an explicit resource cost of searching for non-agricultural employment, which could partly proxy for migration or transition (resource) costs, is unlikely to change our qualitative conclusions. However, since the cost of participation appears to play an important role, evidence on these costs may be critical to ultimately determine the true quantitative effect of policy changes. Relatedly, the benchmark model abstracts from resource losses as a result of inefficiencies in tax collection and in the execution of public investment projects. While reasonable degrees of inefficiency do not change the main message, the degree of convexity of these costs—which may be higher for higher levels of public investment—may ultimately lead to adverse salaried employment effects from higher capital taxes. Put differently, the rise in public investment and its associated effect on private-sector productivity must be large enough to offset the initial adverse effect of higher taxes on private-sector capital accumulation. Our framework does not assume differences in non-agricultural output between salaried and self-employed firms. While consumption taxes may be easier to collect for salariedproduced consumption goods and services, this may not be the case for self-employedproduced consumption goods and services. Assuming that consumption taxes only apply to salaried-produced consumption goods would still affect participation decisions by raising the marginal cost of participation, but the decrease in salaried employment (and increase in self-employment) would likely be larger since consumption taxes would be biased towards salaried-firm output. Despite these caveats, the fundamental mechanisms via which fiscal policy affects sectoral and aggregate employment, unemployment, labor force participation, and macro aggregates would remain intact. be seen as a reduced-form way of capturing the frictions (financial, geographical, etc...) that prevent entry into self-employment.
33
Finally, given data limitations in LICs (and their associated implications for robust model calibrations), we see our policy experiments mainly as a way to quantitatively highlight the relevant transmission mechanisms and implications of fiscal policy in a LIC labor market context rather than uncovering specific policy-related elasticities in a given LIC.
6
Conclusion
Low-income countries (LICs) in Sub-Saharan Africa (SSA) have higher shares of self-employment, employment and agriculture, and informal employment relative to other income groups. They have a limited tax base, revenue that is heavily dependent on external sources, and severe public infrastructure deficits that put a dent on private-sector productivity. Little is known about the labor market and aggregate consequences of moving towards domestic distortionary taxation to raise revenue for public investment in these economies. In this paper, we build a general equilibrium model with labor search frictions and endogenous participation consistent with the employment and tax structure in SSA economies and analyze the labor market and aggregate implications of fiscal reforms aimed at bolstering public investment via domestic taxation. The model features salaried employment, agricultural and non-agricultural self-employment, endogenous labor force participation, and a feedback effect from tax collection to public investment, where the latter positively affects privatesector productivity. Focusing on the main sources of taxation in SSA economies, we analyze the labor market and aggregate impact of individual changes in taxes on non-agricultural consumption and taxes on salaried profits and capital gains. We find that a rise in distortionary salaried capital income taxation can increase salaried employment, total output, and consumption; it reduces non-agricultural self-employment while leaving agricultural self-employment unchanged; and causes a slight increase (decrease) in unemployment (participation). Conversely, a similar rise in consumption taxes reduces salaried employment, output, total consumption, and increases self-employment. The joint inclusion of non-agricultural self-employment, public investment, and endogenous labor force participation for all relevant employment categories—three margins that are generally not jointly present in existing models of LIC labor markets—play an important role for explaining 34
the positive effects of capital tax changes on salaried employment, consumption, and output, as well as the small adverse impact on unemployment. Our results have three main policy implications. First, in economies where self-employment is the most prevalent source of employment, explicitly accounting for endogenous changes in sectoral labor force participation and the full set of available employment options is relevant for characterizing the labor market and aggregate impact of changes in fiscal policy. Abstracting from these two features can lead to considerably different policy outcomes. Second, changes in capital taxes can have positive effects on salaried employment and output so long as government revenue is devoted to public investment, the latter benefits private-sector productivity, and salaried firms can attract individuals working in non-salaried employment into salaried employment. Third, despite its considerable contribution to total employment in SSA economies, self-employment in agriculture plays a limited role as a quantitativelyrelevant adjustment mechanism in response to changes in fiscal policy. Our framework can easily be extended to address important questions related to financial development, firm structure, and structural reforms in a context where heterogeneity in employment arrangements is a critical feature of the labor market, as is the case in SSA economies.
7
Acknowledgements
We thank Margaret McMillan, Kyle Emerick, Kelsey Jack, Sahar Parsa, Adam Storeygard, Adrian Peralta, and participants in the IMF Development Macroeconomics research seminar, the Midwest Macro Meetings Fall 2016, LACEA 2016, and Oxford University’s Center for the Study of African Economies Annual Conference 2017 for useful comments and feedback. This paper is supported by the DFID-IMF Project on Research on Macroeconomics of LowIncome Economies.
35
References [1] Adam, Christopher, David Bevan, and Douglas Gollin. 2013. ”Ruran-Urban Linkages, Transaction Costs, and Poverty Alleviation: The Case of Tanzania,” mimeo. [2] Ag´enor, Pierre-Richard, and Joshua Aizenman. 1999. ”Macroeconomic Adjustment with Segmented Labor Markets,” Journal of Development Economics, Vol. 58, pp. 277-296. [3] Anand, Rahul, Eswar S. Prasad, and Boyang Zhang. 2015. ”What Measure of Inflation Should a Developing Country Central Bank Target?” Journal of Monetary Economics, Vol. 74, pp. 102-116. [4] Arseneau, David M., and Sanjay K. Chugh. 2012. “Tax Smoothing in Frictional Labor Markets,” Journal of Political Economy, Vol. 120(5), pp. 926-985. [5] Baxter, Marianne, and Robert G. King. 1993. ”Fiscal Policy in General Equilibrium,” American Economic Review, Vol. 83(3), pp. 215-334. [6] Berg Andrew, Rafael Portillo, Shu-Chun S. Yang, and Luis-Felipe Zanna. 2013. ”Public Investment in Resource-Abundant Developing Countries,” IMF Economic Review, Vol. 61(1), pp. 92–129. [7] Buffie, Edward, Christopher Adam, Stephen O’Connell, and Catherine Pattillo. 2008. ”Riding the Wave: Monetary Responses to Aid Surges in Low-Income Countries,” European Economic Review, Vol. 52, pp. 1378-1395. [8] Calder´on, Cesar, and Luis Serv´en. 2014. ”Infrastructure, Growth, and Inequality: An Overview,” World Bank Policy Research Working Paper 7034. [9] Charlot, Olivier, Franck Malherbet, and Cristina Terra. 2015. ”Informality in Developing Economies: Regulation and Fiscal Policies,” Journal of Economic Dynamics and Control, Vol. 51, pp. 1-27. [10] Cook, David, and Hiromi Nosaka. 2005. ”Dual Labor Markets and Business Cycles,” Federal Reserve Bank of San Francisco Working Paper 2006-36. [11] De Vreyer, Philippe, and Francois Roubaud. 2013. ”Urban Labor Markets in Sub-Saharan Africa,” The World Bank Group: Washington D.C. http://dx.doi.org/10.1596/978-0-8213-9781-7. [12] Diao, Xinshen, Kenneth Harttgen, and Margaret McMillan. 2016. ”The Changing Structure of Africa’s Economies,” mimeo. [13] Epstein, Brendan, and Alan Finkelstein Shapiro. 2017. ”Employment and Firm Heterogeneity, Capital Allocation, and Countercyclical Labor Market Policies,” Journal of Development Economics, Vol. 127, pp. 25-41. [14] Esfahani, Hadi Salehi, and Mar´ıa Teresa Ram´ırez. 2003. ”Institutions, Infrastructure, and Economic Growth,” Journal of Development Economics, Vol. 70(2), pp. 443-477. 36
[15] Fields, Gary S. 1975. “Rural-Urban Migration, Urban Unemployment and Underemployment, and Job Search Activity in LDC’s,” Journal of Development Economics, Vol. 2, pp. 165-188. [16] Finkelstein Shapiro, Alan. 2014. ”Self-Employment and Business Cycle Persistence: Does the Composition of Employment Matter for Economic Recoveries?” Journal of Economic Dynamics and Control, Vol. 46, pp. 200-218. [17] Finkelstein Shapiro, Alan, and Federico Mandelman. 2016. ”Remittances, Entrepreneurship, and Employment Dynamics Over the Business Cycle,” Journal of International Economics, Vol. 103, pp. 184-199. [18] Gollin, Douglas. 2008. ”Nobody’s Business but My Own: SElf-Employment and Small Enterprise in Economic Development,” Journal of Monetary Economics, Vol. 55(2), pp. 219-233. [19] Gollin, Douglas, Stephen L. Parente, and Richard Rogerson. 2004. ”Farm Work, Home Work and International Productivity Differences,” Review of Economic Dynamics, Vol. 7, pp. 827-850. [20] Gollin, Douglas, and Richard Rogerson. 2013. ”Productivity, Transport Costs and Subsistence Agriculture,” mimeo. [21] Golub, Stephen. 2015. ”Informal Cross-Border Trade and Smuggling in Africa,” Chapter 10, Handbook on Trade and Development, Ed. Oliver Morrissey, Ricardo Lopez, and Kishor Sharma. [22] Golub, Stephen, and Faraz Hayat. 2015. ”Employment, Unemployment and Underemployment in Africa,” Handbook of Africa and Economics, Ed. Justin Lin and Celestin Monga. [23] Grimm, Michael, Jens Kr¨ uger, and Jann Lay. 2011. ”Barriers to Entry and Returns to Capital in Informal Activities: Evidence from Sub-Saharan Africa,” The Review of Income and Wealth, Vol. 57, pp. 25-53. [24] Gutman, Jeffrey, Amadou Sy, and Soumya Chattopadhyay. 2015. ”Financing African Infrastructure: Can the world Deliver?” Global Economy and Development, The Brookings Institution. [25] Harris, John, and Michael P. Todaro. 1970. ”Migration, Unemployment and Development: A Two-Sector Analysis,” American Economic Review, Vol. 60(1), pp. 126-142. [26] Henderson, J. Vernon, Mark Roberts, and Adam Storeygard. 2013. ”Is Urbanization in Sub-Saharan Africa Different?” World Bank Policy Research Working Paper 6481. [27] Hulten, Charles R. 1996. ”Infrastructure Capital and Economic Growth: How Well You Use It May Be More Important Than How Much You Have,” NBER Working Paper No. 5847. 37
[28] IMF. 2014. ”Regional Economic Outlook: Sub-Saharan Africa, Staying the Course,” October 2014, International Monetary Fund: Washington D.C. [29] IMF. 2015a. ”Regional Economic Outlook: Sub-Saharan Africa, Dealing with the Gathering Clouds,” October 2015, International Monetary Fund: Washington D.C. [30] IMF. 2015b. ”Fiscal Policy and Long-Term Growth,” IMF Policy Paper June 2015. [31] Kerr, Andrew. 2012. ”A Model of Comparative Advantage with Matching in the Urban Tanzanian Labour Market,” CSAE Working Paper WPS/2012-21. [32] Kumar, Alok, and Herbert J. Schuetze. 2007. ”Self-Employment and Labor Market Policies,” mimeo. [33] Melina, Giovanni, Shu-Chun S. Yang, and Luis-Felipe Zanna. 2015. ”Debt Sustainability, Public Investment, and Natural Resources in Developing Countries,” Economic Modelling (forthcoming). [34] OECD. 2009. “Overview: Data on Informal Employment and Self-Employment,” in Is Informal Normal? Towards More and Better Jobs in Developing Countries, OECD Development Centre, Ed. Johannes P. J¨ utting and Juan R. de Laiglesia. [35] Pritchett, Lant. “The Tyranny of Concepts: CUDIE (Cumulated, Depreciated, Investment Effort) Is Not Capital,” Journal of Economic Growth, Vol. 5, pp. 361-384. [36] Restuccia, Diego, Dennis Tao Yang, and Xiaodong Zhu. 2008. ”Agriculture and Aggregate Productivity: A Quantitative Cross-Country Analysis,” Journal of Monetary Economics, Vol. 55(2), pp. 234-250. [37] Rud, Juan Pablo, and Ija Trapeznikova. 2016. ”Wage Dispersion, Job Creation and Development: Evidence from Sub-Saharan Africa,” mimeo. [38] Satchi, Mathan, and Jonathan Temple 2009. ”Labor Markets and Productivity in Developing Countries,” Review of Economic Dynamics, Vol. 12(1), pp. 183-204. [39] Admasu Shiferaw, and Arjun Bedi. 2009. ”The Dynamics of Job Creation and Job Destruction: Is Sub-Saharan Africa Different?” IZA Discussion Paper No. 4623. [40] Ulyssea, Gabriel. 2010. ”Regulation of Entry, Labor Market Institutions and the Informal Sector,” Journal of Development Economics, Vol. 91(1), pp. 87-99. [41] Yepes, Tito, Justin Pierce, and Vivien Foster. 2009. ”Making Sense of Africa’s Infrastructure Endowment: A Benchmarking Approach,” World Bank Policy Research Working Paper 4912. [42] Zenou, Yves. 2011. ”Search, Migration, and Urban Land Use: The Case of Transportation Policies,” Journal of Development Economics, Vol. 96, pp. 174-187.
38
